When managing a filling station chain, it is necessary to solve the following tasks in dynamics: determining the volume of purchases by types of oil products sold and redistributing the available volume of various types of oil products to filling station chain. The peculiarity of this control task is that the replenishment of the storage of oil products is carried out centrally, and the sale of each type of fuel is carried out at separate filling stations of the network, that is, in many points. At the same time, the volume of each type of fuel sold by each station of the network should be taken into account separately, taking into account the seasonal demand for a specific oil product. In addition to modeling demand, information is needed on the volume of petroleum products that can be purchased from suppliers with the possibility of increasing or decreasing the volume of purchases due to demand. To solve this problem, a list of suppliers of a specific type of oil product with a possible range of supply volumes must be determined. In the proposed model, an attempt is made to solve the above problems. To solve them, a management model was developed, which was implemented using the Visual Studio C # programming environment and MS SQLServer DBMS. When developing the structure of the database tables, the task of managing a network of filling stations based on the Petri net was taken into account: that is, the database tables provided for storing information on the volumes of supplies of petroleum products, the volume of sales of each type of petroleum product, as well as the time of deliveries and sales. To solve the problem of centralized procurement for all types of fuel, the database provided tables with information about suppliers and possible volumes of supplies by them of various types of petroleum products. To solve the problems of forecasting demand, the model includes algorithms for predicting the volume of sales of petroleum products based on the accumulated time series of data for each type of fuel separately. The forecasting was carried out in order to assess the required volumes of purchases of oil products for the coming period. The prediction algorithms are implemented using two methods: linear approximation and exponential smoothing. Both algorithms take into account the seasonality of demand.